A parameter estimation method based on aggregated data

Abstract

In constructing multiprocessor-based distributed process control systems, one approach is to use low-end processors to carry out direct control tasks at the local layer, while leaving more sophisticated tasks for high-end processors at the supervisory layer. For such a system architecture, reducing the data transmission on the communication link is a challenging problem. Hung and Lefkowitz proposed a data aggregation algorithm for compressing the original input/output data at the local layer before passing them to the supervisory layer. At the supervisory layer, however, there is a lack of an effective algorithm to perform the parameter estimation task based on the aggregated data. This paper describes a method which significantly improves the effectiveness of this task. Systems under control are assumed to be linear with distinct eigenvalues. Simulation results are included.